Use of satellite radiances in the global assimilation system at JMA

Transcription

1 Use of satellite radiances in the global assimilation system at JMA Kozo Okamoto, Hiromi Owada, Yoshiaki Sato, Toshiyuki Ishibashi Japan Meteorological Agency ITSC-XV: Maratea, Italy, 4-10 October 2006

2 Satellite data assimilated in JMA operational system Radiances NOAA-15,16/AMSU-A Aqua/AMSU-A NOAA-15,16,17/AMSU-B DMSP-13,14/SSMI TRMM/TMI Aqua/AMSR-E Other satellite data ATOVS radiances MWR radiances AMV from MTSAT1R, GOES-11,12Meteosat-5,8 AMV from Aqua+Terra/MODIS sea surface winds from QuikSCAT/SeaWinds Under development CSR (Clear Sky Radiance) from MTSAT-1R refractivity index from GPS occultation radiances from Aqua/AIRS CSR

3 JMA global assimilation system 4DVar incremental method outer: TL319L40, inner: T106L40 model top : 0.4 hpa Background error covariance based on NMC method Physical processes included are simplified versions of gravity drag, long wave radiation, convection, cloud condensation. VarBC (Variational Bias Correction) assimilation window: 6h 1h time slot X 6 cut off time: 2h20m for early anal, 5h35m- 11h35m for final anal

4 VarBC (1/2) adaptive bias correction scheme incorporated in a variational scheme (Derber and Wu 1998; Dee 2004) bias correction coefficients are analyzed as analysis variables J(x)=(x-x b ) T B -1 (x-x b )+[y-h (x)] T R -1 [y-h (x)]+(β- β b ) B β -1 (β- β b ) H '( x) = H ( x) + βi pi xb β :BC coef., pi : i= 1 obtain balanced coefficients among targeted obs, model, and other obs BC coeffs obtained from only guess could be contaminated by model biases easy to adjust to model and data changes BC coef time-sequence monitor may be used as instrumental quality monitor applied to all radiance data predictors for AMSU-A/B: ILR, Ts, TCCLW, 1/cosθconst predictors for MWR: TCWV, Ts, Ts2, wind.speed, 1/cosθconst N ( ) predictors ILR: integrated (weighted) lapse rate TCCLW: total column cloud liquid water retrieved by AMSU-A TCWV: total column water vapor retrieved by MWR

5 VarBC (2/2) BC coef variation monitor may be used as instrumental quality monitor ex. NOAA15 AMSU-A ch6 Fluctuation of VarBC coef well correspond to instrumental temperature fall Instrumental Temp Mar /1/1 2006/1/8 2006/1/ /1/ /1/ /2/5 2006/2/ /2/ /2/ /3/5 2006/3/ /3/ /3/26 Apr 2005 Jan O-B Bias & RMSE O-B Bias & RMSE 06/2/1 06/2/8 06/2/15 06/2/22 06/3/1 06/3/8 06/3/15 06/3/22 06/3/29 w/ovarbc-bias w/ovarbc-rmse wvarbc-bias wvarbc-rmse

6 ATOVS radiance assimilation NOAA-15,16,17,Aqua/AMSU-A,-B level-1c AMSU-A ch.4-13 and AMSU-B ch.3-5 QC less cloudy-affected radiances are used depending on surface conditions thinning AMSU-A:250km, -B:180km choose one satellite among overlapped satellites in each 1-h time slot BC 2-step scheme: scanbc + airmassbc scanbc removes predefined O-B average at each scan position airmassbc removes the residual using VarBC

7 A9 O-B w/o BC A10 Bias Corr A9 O-A w VarBC A10 A11 A12 monthly average O-B & O-A A11 A12 A13 AMSU-A9-13 NOAA16 Aug2004 A13 A9 O-B w scanbc A10 A9 O-B w VarBC A10 A11 A12 A11 A12 A13 A13

8 ATOVS assimilation changes in Aug2006 improve QC adopt MSPPS latest version for MW-cloud detection stricter gross error QC, remove edge scans recalculate scanbc change VarBC predictors modify obs errors of AMSU-A reduce obs error inflation factor, 2.3 to 1.2 obs errors are inflated in 4DVar main analysis to complement neglecting horizontal error correlation and balance among contributions from other observations and guess. O-B has been getting smaller due to using level-1c data, revising scanbc and including VarBC inflation inflation factor factor = = Z ANC Z ANC inflation inflation factor factor = = Z ANC Z ANC NH SH NH SH

9 ATOVS assimilation changes in Aug2006 impacts on forecasts winning percentages : (test-cntl)/all Psea T850 Z500 Wsp850 Wsp250 Improvement rate wrt RMSE : (cntl-test)/cntl ANC diff : test-cntl average of 1 to 9 day forecasts: cntl-test better neutral worse

10 MWR radiance assimilation (1/2) DMSP/SSMI, TRMM/TMI, Aqua/AMSR-E less cloud-affected radiances over the ocean with SST > 5 deg.c assimilate only vertical polarized channels at GHz not change surface variables through emissivity Jacobian VarBC corrects biases against model Comparison with TRMM 3-day-averaged TCPW MWR radiance assimilation leads to better representation of TCPW Global Analysis Global Analysis w/o MWR radiance w/o MWR radiance TRMM difference Apr 5-May 10-May 15-May 20-May 25-May BIAS RM SE 30-May TRMM difference w MWR radiance w MWR radiance Start MWR radiance assimilation 5 May 25 May

11 MWR radiance assimilation (2/2) Impacts on forecasts better precipitation Correlations of 1day-forecast against GPCP: w/omwr:0.881 vs. wmwr:0.891 (Aug2004) w/omwr:0.835 vs. wmwr:0.841 (Jan2004) better typhoon track smaller errors in Psea and wind speed at both 850 and 250 hpa operational in May 2006 Position error [km] wmwr rad w/omwr rad Forecast time (00-90h)

12 MTSAT-1R CSR assimilation (1/2) clear sky radiances (CSR) from MTSAT-1R WV channel Meteorological Satellite Center (MSC) of JMA produces ready for dissemination to NWP community clear ratio 10 O-B map UTC Sep O-B mean=-1.7 std=2.18

13 MTSAT-1R CSR assimilation (2/2) under development improve forecast skills, including 500T, in the Tropics and summer hemisphere However, forecast skills in the winter hemisphere are degraded may conflict with AMV (biases)? revising thinning interval (now every 1-h time slot) observation errors (now 1.5K) (Cntl-Test)/Cntl Mean, Mean, Std Stdand RMSE RMSE of of O-B O-B wbc wbc O-B O-B wbc wbc (Cntl-Test)/Cntl in in zonal zonal mean mean RMSE RMSE of of Z Z at at FT=5d FT=5d Aug.2005 Aug /20 7/30 8/9 8/19 8/29 90N 90S

14 Plans ATOVS improve MW emissivity to use more data over land/snow/sea-ice add NOAA18 use AP-RARS data in early analysis CSR from geostationary satellites AIRS and IASI radiances SSMIS radiances (Global DAS) and retrievals (Meso-scale DAS) GPS occultation as refractive index ASCAT (ambiguity) winds cloud/rain-affected radiances of MW and IR sensors

15 AP-RARS (Asia-Pacific Regional ATOVS Retransmission Service) 10 stations from Japan(2), China(3), Australia(4), Korea(1). Coverage of the data received at Syowa Station, Antarctica Coverage of the data received at MSC Shaded area shows coverage of the data received at AP-RARS stations (as of Sep. 2006).

16 Impacts of AP-RARS expected to improve early analysis by adding early delivery data shrink the difference from data-rich final analysis comparison of early analysis with final analysis w and w/o AP- RARS Z30 at 00UTC on Jul 18, 2006 Early.anal Early.anal Final.anal Final.anal w/o w/o AP-RARS AP-RARS Early.anal Early.anal Final.anal Final.anal w AP-RARS AP-RARS data distribution of AP-RARS (NOAA16 AMSU-Ach1)

17 Supplemental slides

18 NWP operational system (as of Sep.2006) Global Model (GSM) Regional Model (RSM) Typhoon Model (TYM) Mesoscale Model (MSM) One-week Ensemble Objectives Medium-range forecast Short-range forecast Typhoon forecast Disaster reduction One- week forecast Forecast domain Global East Asia Typhoon and its surrounding Japan and its surrounding Global Grid size / Number of grids deg 640 x 320 (TL319) 20 km 325 x km 271 x km 721 x deg 320 x 160 (TL159) Vertical levels / Top hpa hpa hpa 50 21,800m hpa Forecast hours (Initial time) 90 hours (00 UTC) 216 hours (12 UTC) 36 hours (06, 18 UTC) 51 hours (00, 12UTC) 84 hours (00, 06, 12, 18 UTC) 15 hours (00, 03, 06, 09, 12, 15, 18, 21 UTC) 9 days (12 UTC) 51 members Analysis 4D-Var 4D-Var Interpolated from Global Analysis 4D-Var Global Analysis with ensemble perturbations

19 super computer at JMA Satellite Data Processing Numerical Weather Prediction Mar Mar nodes HITACHI SR11000J1 50nodes 6.1Tflops =1.9 GFlops x 16 processors x 50 nodes x 1 80nodes HITACHI SR11000K1 160nodes 21.5Tflops =2.1 GFlops x 16 processors x 80 nodes x 2 80nodes 13.1TB Main memory 36.2TB Disk 2.0PB Tape

20 ATOVS VarBC predictors ILR, Ts, TCCLW, 1/cosθconst dependence on the predictors TCCLW dependence for AMSU-A4 and lower tropospheric ch. Separation of ocean and land may be needed for Ts unclear dependence for AMSU-B AMSU-A4 AMSU-A4 AMSU-A8 AMSU-A8 TCCLW TCCLW predictors vs. O-B-scanBC ILR ILR Ts Ts